Hi, >I am using VirtualBox (from the Debian repositories) for virtualisation. One of >my VMs uses two virtual hard drives (Fixed-size VDIs of 120 GBytes each, >visible to the guest as SATA drives). I am trying to delete the oldest snapshot >on this VM, so the "original" (fixed-size) VDI files are to be modified as >changes are merged back into them from the oldest "difference" VDIs. Snapshot >deletion was triggered using the "delete snaphot" button in the QT GUI; the >host system was freshly booted, and no VMs were running. >VirtualBox merges the "difference" VDI back into the first fixed-size VDI. It >then deletes the "difference" VDI from the host filesystem, but instead of >working on the second virtual disk, the progress indicator just disappears. The >snapshot is still visible; the Virtual media Manager lists the "difference" VDI >that was just being deleted as "missing" (so obviously, the media registry was >not updated). >dmesg delivers the message "[ 868.272600] DeleteSnap[2018]: segfault at 31 ip >000000000052bf8c sp 00007fb4b7b6d9c0 error 4 in VBoxSVC[400000+482000]" (the >precediung and following lines are not associated with VirtualBox). I have no >hints of physical media errors in the kernel log, so I consider the hard drive >as OK here. >htop shows that I am now having a total of three active VirtualBox processes, >which load two physical CPU cores at 100% each. > >The expected behaviour would have been that nothing crashes, the VirtualBox >media registry is updated correctly, and the second virtual disk is processed >as well. Then, the configuration file for this VM shall be updated according to >the snapshot deletion, and the snapshot shall disappear from the snapshot tree. > >I restored the VM from a backup and tried again, with the same result (however, >I did not check dmesg and htop in the 2nd attempt). > >The last time I deleted snapshots for this VM was with VirtualBox 5.0.24; I did >not have trouble there. VirtualBox 5.1.6 now shows the described behaviour. > >I have no problems with deleting snapshots for VMs that have only one virtual >hard drive in VirtualBox 5.1.6, so this seems to be linked to the number of >virtual drives.
can you please open an upstream ticket [1] and link it here? I pinged upstream on irc, and they said they already got a similar email with a similar bug report. Having a ticket will allow me to see how/when the bug is fixed and close it eventually thanks! G. [1] https://www.virtualbox.org/newticket